The True Costs of Participatory Sanitation: Evidence from Community-Led Total Sanitation Studies in Ghana and Ethiopia

Science of the Total Environment 601–602 (2017) 1075–1083

Contents lists available at ScienceDirect

Science of the Total Environment

journal homepage: www.elsevier.com/locate/scitotenv

The true costs of participatory sanitation: Evidence from community-led total sanitation studies in Ghana and Ethiopia

Jonny Crocker a,⁎, Darren Saywell b, Katherine F. Shields a,PeteKolskya, Jamie Bartram a a The Water Institute, University of North Carolina at Chapel Hill, 148 Rosenau Hall, CB #7431, Chapel Hill, NC 27599-7431, USA b Plan International USA, 1255 23rd Swt NW Suite 300, Washington, DC 20037, USA

HIGHLIGHTS GRAPHICAL ABSTRACT

• Bottom-up cost analysis of community- led total sanitation in Ghana and Ethiopia • The program cost was $30–82 per household targeted in Ghana, and $14–19 in Ethiopia. • Local investments were $8–22 per household targeted in Ghana, and $2–3 in Ethiopia. • The ﬁndings are relevant for policy, prac- tice, and further research.

article info abstract

Article history: Evidence on sanitation and hygiene program costs is used for many purposes. The few studies that report costs Received 27 January 2017 use top-down costing methods that are inaccurate and inappropriate. Community-led total sanitation (CLTS) is Received in revised form 25 May 2017 a participatory behavior-change approach that presents difficulties for cost analysis. We used implementation Accepted 31 May 2017 tracking and bottom-up, activity-based costing to assess the process, program costs, and local investments for Available online xxxx four CLTS interventions in Ghana and Ethiopia. Data collection included implementation checklists, surveys, fi Editor: Simon Pollard and nancial records review. Financial costs and value-of-time spent on CLTS by different actors were assessed. Results are disaggregated by intervention, cost category, actor, geographic area, and project month. The average Keywords: household size was 4.0 people in Ghana, and 5.8 people in Ethiopia. The program cost of CLTS was $30.34–$81.56 Cost per household targeted in Ghana, and $14.15–$19.21 in Ethiopia. Most program costs were from training for Process three of four interventions. Local investments ranged from $7.93–$22.36 per household targeted in Ghana, and Behavior-change $2.35–$3.41 in Ethiopia. This is the first study to present comprehensive, disaggregated costs of a sanitation Sanitation and hygiene behavior-change intervention. The findings can be used to inform policy and finance decisions, Hygiene plan program scale-up, perform cost-effectiveness and benefit studies, and compare different interventions. CLTS The costing method is applicable to other public health behavior-change programs. © 2017 The Authors. Published by Elsevier B.V. This is an open access article under the CC BY license (http:// creativecommons.org/licenses/by/4.0/).

1. Introduction

Cost evidence informs policies, program design and scale-up, and research. Such evidence is lacking for water, sanitation, and hygiene ⁎ Corresponding author. (WaSH) programs that are participatory, involve capacity building, or E-mail address: [email protected] (J. Crocker). target behavior-change. Improving this evidence is a priority, as

http://dx.doi.org/10.1016/j.scitotenv.2017.05.279 0048-9697/© 2017 The Authors. Published by Elsevier B.V. This is an open access article under the CC BY license (http://creativecommons.org/licenses/by/4.0/). 1076 J. Crocker et al. / Science of the Total Environment 601–602 (2017) 1075–1083 meeting the Sustainable Development Goals (SDGs) will necessitate The few studies with primary cost data for sanitation and hygiene both a scale-up of efforts to meet universal targets, and a shift in the behavior-change omit management and other software costs, use means of implementation toward capacity building, local participation, broad assumptions to fill data gaps, rely on recall by few respondents, and behaviors (UN General Assembly, 2015). and sample non-representative respondents, all problems that the WaSH and other public health programs have characteristics that authors acknowledge (Borghi et al., 2002; Briceño and Chase, 2015; make costing difficult: complex institutional arrangements, cross- Burr and Fonseca, 2011; Evans et al., 2009; Robinson, 2005; Trémolet subsidies, flexible implementation, and local investments. Complex et al., 2010). Importantly, these studies all use TDC methods. One institutional arrangements spread costs across organizations, resulting study presents a BUC analysis of a WaSH program (Briceño and Chase, in inconsistent and incomplete financial tracking. Cross-subsidies arise 2015). However, data were non-representative, costs were not disag- when programs share resources (such as vehicles or training). Participa- gregated beyond program and household, and some local investments tory, behavior-change programs are inherently flexible, extensively were omitted. Another study presents costs and benefits of latrine con- adapted, and field activities often do not match workplans or struction (Dickinson et al., 2015), though they focused on household budgets. Local actors and communities contributing time or money costs. (“local investments”) is common in participatory behavior-change We performed a BUC process and cost analysis of four CLTS interven- programs. tions in Ghana and Ethiopia. We chose to present costs per population Community-led total sanitation (CLTS) epitomizes these costing targeted rather than per population reached to focus this paper on our challenges. CLTS is a participatory approach in which facilitators visit costing methods, findings, and implications. Converting to cost per pop- villages and trigger awareness of sanitation issues during a community ulation reached adds another layer of complexity, as there are multiple meeting. Facilitators then perform follow-up visits to villages to gener- reasonable outcomes that can be used for this conversion, and the num- ate a community-wide effort to become open defecation free (ODF). ber of households reaching any given outcome depends on context. Eliminating open defecation is included in the SDGs (UN General However, it is fine to convert to cost per household reached, which is Assembly, 2015), as open defecation can cause malnutrition (Dangour done by dividing by the program costs by the percent of program house- et al., 2013), child stunting (Spears, 2013), and death (Prüss-Ustün holds that reached the desired outcome. et al., 2014). CLTS has spread to over 60 countries since 2000, in part be- We disaggregated results by intervention, geographic area, actor, cause of perceived low cost: it rarely includes subsidized latrines time, and cost category to enable assessment of what drives variability, (Institute of Development Studies, 2016; Kar and Chambers, 2008). and how costs would transfer to other programs and settings. This study Costing methods are either top-down or bottom-up. Top-down cost- was implementation research conducted by Plan International USA and ing (TDC) involves dividing a program's budget or total expenditures by The Water Institute at UNC. the number of units (villages, households, individuals) targeted or reached. It is appealing due to its use of minimal, routinely collected data (budgets, expenditures, population targeted or reached), simple 2. Methods analysis, and that it can be retrospective. TDC can be accurate when: budget and expenditures represent all program costs, and only that pro- 2.1. Program description gram's costs, and the population served is unambiguous; conditions un- common for WaSH programs. When cross-subsidies exist, TDC will The four interventions were: in Ghana, (1) NGO-facilitated CLTS, and under- or over-estimate costs depending on which program purchased (2) NGO-facilitated CLTS with additional training for natural leaders; shared resources. Neglecting local investments leads to underestimated and in Ethiopia, (3) health extension worker (HEW) and kebele total costs, leaves potentially disadvantaged beneficiaries out of cost leader-facilitated CLTS, and (4) teacher-facilitated CLTS. Natural leaders considerations, and contributes to poorly informed policy (Garber and are motivated community members who encourage others to construct Phelps, 1997). In the absence of these conditions, one advantage of and use latrines. A kebele is the lowest administrative unit in Ethiopia, TDC is that it can capture management and overhead costs better than comprising approximately 20–30 villages and 5000 people in rural bottom-up costing (BUC). TDC does not allow disaggregation of costs areas. Implementation activities, actors, and timeline for the four by category (e.g. management, training, hardware), actor (e.g. govern- interventions analyzed here are in the supplement. Project evaluations ment, non-governmental organization, community), time (e.g. by and implementation narratives are presented elsewhere (Crocker month), or by project or setting (Chapko et al., 2009). It is inappropriate et al., 2017, 2016a, 2016b; Plan International Ethiopia, 2015; Plan Inter- when these factors are of interest, as is frequent in WaSH. national Ghana, 2015). Facilitation has three stages: pre-triggering - BUC involves careful tracking and analysis of implementation to cal- building a rapport and buy-in with community members; triggering - culate costs and assign them to activities. Regular program activities meeting with communities to conduct group activities that elicit (timesheets, household surveys) can be adapted to collect the data emotional reactions, such as shame and disgust, to generate motivation needed for BUC. However, many cost analyses are done retrospectively, to eliminate OD; and follow-up - monitoring a community's progress which precludes BUC. Additionally, the analysis is time consuming, and guiding them toward eliminating OD. Further details on the CLTS complex, and expensive (Carey and Burgess, 2000), which could explain approach can be found in the CLTS Handbook (Kar and Chambers, BUC's scarcity. BUC is more appropriate than TDC for the complexity of 2008). WaSH. BUC overcomes the main sources of error and bias (Adam et al., For all four interventions, implementation began with an orientation 2003), and enables analysis of variation in cost, economies of scale, and workshop for district government officials. For intervention 1 (Ghana), comparison of interventions (Chapko et al., 2009), which are valuable implementation proceeded with CLTS facilitation by Plan International for program design and management. and local NGO (LNGO) staff (Table 1) with no formal training of local ac- While there is a growing body of evidence on the effectiveness of tors. Henceforth, Plan International and their contracted LNGOs are re- sanitation interventions (Garn et al., 2017), cost evidence is lacking. ferred to as “Plan”. Intervention 2 (Ghana) included all the activities of Several authors have compiled secondary data to model the costs and intervention 1, with the addition of Plan training natural leaders to sup- benefits of achieving global WaSH targets (Haller et al., 2007; Hutton port CLTS. For interventions 3 and 4 (Ethiopia), Plan trained kebele and Varughese, 2016), or to compare different interventions (Hutton leaders, and either HEWs or teachers, as facilitators. LNGOs were not et al., 2007; Whittington et al., 2012). They emphasize that low quality contracted in Ethiopia. The four CLTS interventions cover a range of im- or lacking cost evidence forces assumptions and excluding cost catego- plementation arrangements and modalities as practiced in other organi- ries, resulting in incomplete and potentially misleading results. There is zations and countries (Venkataramanan, 2016, 2012), so the findings also lacking evidence for water supply programs (Hunter et al., 2009). are relevant beyond this project. J. Crocker et al. / Science of the Total Environment 601–602 (2017) 1075–1083 1077

Table 1 Table 2 Plan implementation activities for four CLTS interventions in Ghana and Ethiopia. Cost categories and components.

Category Activity Ghana Ethiopia Category Components

NGO NGO CLTS + NL HEW Teacher Program costs Management Paid time – manager CLTS training CLTS CLTS Paid time – field staff Office rent Management Project management •• •• Office supplies Training District government •• •• Traininga Paid time – trainers orientation Transportation Training kebele leaders •• Venue, accommodation, meals Training HEWs • Per-diems Training teachers • Facilitationb Paid time – Plan staff Training natural leaders ••• Paid time – government officials Facilitation Facilitation •• •• Transportation Monitoring •• •• Per-diems ODF celebration •• •• ODF celebration costs Abbreviations: NGO, non-governmental organization; CLTS, community-led total sanita- Local investments Local actor time Unpaid time – during training tion; NL, natural leader; HEW, health extension worker; ODF, open-defecation free. Unpaid time – traveling to training Unpaid time – during Plan's visits Unpaid time – in Plan's absence 2.2. Context Community activity Unpaid time – during Plan's visits Unpaid time – in Plan's absence Pre-existing factors in Ghana and Ethiopia enabled the interven- Unpaid time – latrine construction c tions: national government had included CLTS in policy and established Latrine spending Hired labor Purchased materials support mechanisms such as coordinating committees, and Plan had prior experience implementing CLTS and partnering with national and a All activities outside villages that included local actors or community members. In- sub-national government. This study does not include costs associated cludes orientations, training, and review meetings. b Costs borne by Plan for activities within villages and kebeles. with building government support for CLTS. Country economics also af- c Excludes spending on communal latrines. fect costs. Gross domestic product (GDP) per capita in 2015 was $1370 in Ghana and $619 in Ethiopia (The World Bank, 2017). Minimum wage at the project outset was $3.19 per day in Ghana, and $1.07 per day in Paid time was calculated by multiplying hours spent on CLTS by Ethiopia. hourly pay. Time in training and facilitation was aggregated from checklists, and allocated to intervention, region, actor, and month using meta- data. Travel time was estimated using checklist data, Google Earth, and 2.3. Data collection and management discussions with Plan staff. Management time was estimated from a checklist given to Plan staff after the interventions ended. When actors We developed new data collection tools to track implementation were not paid hourly, a 50-week/2000-hour work-year, 40-hour work- activities from the national- to village-level and to estimate local actor week, and 8-hour work-day were assumed. and community member activity, including checklists for management Plan transportation costs were estimated using the American Auto- activities, training, and facilitation, and surveys for local actors and mobile Association (AAA) guidelines (AAA, 2015), using intervention- households (in the supplement). Checklists were developed by UNC re- and study site-specific parameters (details in supplement). Plan reim- searchers and Plan field staff, built on Plan's previous tools, and were de- bursed trainees at a flat rate for transportation, which was used to calcu- signed to be simple and quick to fill out to maximize compliance and late their transport costs. consistency. Costs for office rent and supplies, training venue rental, and training Local actors and households were surveyed on their interactions, ac- materials were extracted from financial records, and allocated based on tivities, and latrine spending. These surveys, which were also used to implementation activities. Unit costs for accommodation and meals for evaluate the outcomes of the interventions, are described in more detail training attendees, and per-diems for Plan and government staff during in prior publications (Crocker et al., 2016a, 2016b). Discussions with village visits, were multiplied by person-days in training and in the field. Plan staff three times per year clarified details and provided additional Spending on latrine construction (hired labor and purchased mate- data. Unit costs including staff salaries, vehicle purchases, training rials) was self-reported in household surveys. In Ghana, household- venue rental, accommodation and meals, per-diems, and district reported latrine age was used to identify latrines built during the CLTS government contracts were extracted from Plan's quarterly financial re- interventions; in Ethiopia, baseline survey data was used. Since latrine ports and discussions with staff. Web resources and literature were age was not measured in Ethiopia, total latrine spending was distributed reviewed for general parameters such as official exchange rates and evenly across project months. national minimum wages. Unpaid time was monetized using estimated value-of-time for each Checklist data were entered into Microsoft Access 2013, and checked actor. Unpaid time does not represent a financial investment; however, for errors and gaps, which were corrected through correspondence with we included it as it is necessary for implementing CLTS, and some con- Plan staff. Surveys were analyzed in STATA SE13. Costs were calculated sider it a software cost (Whittington et al., 2012). Local actor and com- in Microsoft Excel 2013. munity member time engaged in CLTS activities and constructing latrines (including pit-digging) was calculated from checklists filled 2.4. Analysis out by Plan when they were present, and otherwise estimated from survey responses. Local government, HEW, and teacher wages were used Costs were categorized as program costs (management, training, fa- as their value-of-time, since they were fully employed. The national cilitation) and local investments (local actor time, community member minimum wage and a value-of-time to minimum wage ratio of 0.5 time, latrine spending). These categories were further split into compo- were used for natural leaders and community members. nents that could be calculated or estimated (Table 2). “Local invest- A sensitivity analysis for each estimated parameter was performed ments” are those made by local actors, which includes sub-national by calculating the change in cost associated with a ±50% change in pa- government and natural leaders in both Ghana and Ethiopia; and kebele rameter values. The full analysis framework, unit costs and sources, and leaders, teachers, and HEWs in just Ethiopia. sensitivity analysis are in the supplement. 1078 J. Crocker et al. / Science of the Total Environment 601–602 (2017) 1075–1083

3. Results The aggregate local investment in Ethiopia was $3.41 per household targeted in villages receiving HEW-facilitated CLTS, and a lower $2.35 Population numbers and implementation details are in Table 3.In where teachers facilitated. The difference was due to teacher- Ghana, Plan's efforts focused on facilitating CLTS themselves, which facilitated CLTS being associated with lower attendance at community involved frequent visits to project villages, whereas in Ethiopia Plan's ef- meetings, and fewer households constructing latrines (Table 3). Expen- forts focused on training local actors as facilitators and thus Plan staff ditures on latrines per household targeted were over 30 times higher in spent less time in villages. The program costs and local investments Ghana than Ethiopia. This dramatic difference was mostly due to that follow are presented per household targeted. households in Ghana purchasing materials for latrine construction (e.g. cement, wood, PVC) while nearly all households in Ethiopia built latrines from free materials (e.g. sticks, mud), and partly due to 3.1. Program costs households in Ghana hiring labor to help with latrine construction, while no households in Ethiopia hired labor. Additional details on Program costs (Fig. 1) are broken into management, training, and latrine construction and other outcomes are reported in prior facilitation. In Ghana, NGO-facilitated CLTS cost $30.34 per household, publications. rising to $81.56 when natural leader training was added. In Ethiopia, In Ghana, hired labor for latrine construction was approximately HEW-facilitated CLTS cost $19.21 per household targeted, dropping to one-quarter and purchased latrine materials approximately one-half $14.15 for teacher-facilitated CLTS. of local investments (Table 5). In Ethiopia, where household spending Training was 4% of the program cost for NGO-facilitated CLTS in on latrines was much lower, the value-of-time for local actors and com- Ghana (training district government at 1-day orientation meetings), munity members exceeded 80% of local investments. rising to 58% of program cost where natural leaders were trained Sensitivity analysis of estimated parameters revealed that local in- (Table 4). Most of program cost in Ethiopia was training, where local vestments were most sensitive to changes in value-of-time for local ac- actors were trained as facilitators. Further disaggregated costs (by re- tors. Changing value-of-time estimates by ±50% results in a change in gion, intervention, and month), and normalized per-intervention and local investments of up to 9.7% in Ghana, and 42.9% in Ethiopia. per-village, are in the supplement. The cost of training Plan staff is not included here, as in both countries they were already trained in CLTS 3.3. Time contributions to CLTS at the project outset. Salary, transport, accommodation and meals, and rent and In villages in Ghana that received only NGO-facilitated CLTS (the one purchased materials contributions to cost categories are presented in intervention with no local actor training), for each hour that Plan spent Fig. 2. Management costs are split between salaries and ofﬁce expenses. on CLTS, local actors contributed 0.5 h, and community members con- The cost of training local actors is dominated by accommodation and tributed 5.9 h (Table 6). In villages in Ghana where Plan trained natural meals for the trainers and trainees, followed by transportation. Trans- leaders, both local actor and community member time on CLTS in- portation is the principal facilitation cost, followed by salaries. creased (to 2.5 and 7.5 h for each hour of Plan's time). Sensitivity analysis of estimated parameters revealed that program In Ethiopia, for every hour of Plan's time, local actors contributed costs are most sensitive to changes in travel time to project villages. 4.7–6.8 h to CLTS—higher than in Ghana (Table 6). Community mem- Changing travel times by ±50% results in up to a 12.4% change in pro- bers contributed 27–28 h to CLTS per hour spent by Plan—over triple gram cost in Ghana, and 5.4% in Ethiopia. the ratio in Ghana. Plan staff in Ethiopia spent far more time training local actors than they did facilitating within villages, in contrast to 3.2. Local investments Ghana where Plan's focus was on facilitation within villages. Plan's focus on training in Ethiopia resulted in Plan spending less time on The ﬁnancial costs (hired labor and purchased hardware) and un- CLTS per 10,000 people targeted than in Ghana. paid value-of-time for local actors and community members are Individually, trained local actors contributed 1.1–4.6 h per-week to shown in Fig. 3. The aggregate local investment in Ghana was $7.93 CLTS on average (Table 7). Kebele leaders in Ethiopia were the most ac- per household targeted in villages receiving NGO-facilitated CLTS, and tive, contributing up to 4.6 h per week or 12% full-time equivalent (FTE) a substantially higher $22.36 in villages where natural leaders were to CLTS. On average, community members spent far less time than local trained, due both to more households building latrines (Table 3), and actors on CLTS (under 6 min per week individually). However, this be- higher spending per latrine. lies the fact that not every community member spends time on CLTS. Community members also form the biggest group of actors, who collectively spent the most time on CLTS (Table 6). In Ghana, Plan's time spent on CLTS declined after 12 months (Fig. 4), Table 3 Descriptive statistics for villages receiving four CLTS interventions. when LNGO contracts ended and the number of facilitators dropped from 16 to 4. Local actor time peaked during training in months 6 and Variable Ghana Ethiopia 15. Community activity peaked during triggering in months 3–5. The NGO NGO CLTS + NL HEW Teacher peak at month 8 is because many households reported constructing CLTS training CLTS CLTS their latrine “one year ago” on a survey in month 20, thus activity Regions 3 3 2 2 shown in month 8 likely occurred across multiple months. Kebeles –– 24 In Ethiopia, Plan's time spent on CLTS was much lower than in Ghana, Villages 29 29 54 111 and was more evenly distributed. Plan and local actor activity peaked dur- Households 3443 3312 1624 3838 Population 14,269 12,936 9829 21,724 ing training in months 2, 3, and 13. Community activity peaked during Average household size (people) 4.1 3.9 6.1 5.7 triggering in month 4, and during ODF celebrations in months 9–11. Total Plan village visits 350 375 11 22 –– Kebele leaders, HEWs, and 20 76 4. Discussion teachers trained Natural leaders trained 0 230 51 113 Households building latrines 7% 15% 21% 17% 4.1. Program costs during CLTS

Abbreviations: NGO, non-governmental organization; CLTS, community-led total sanita- We present costs per household targeted in this paper. Cost per tion; NL, natural leader; HEW, health extension worker. household reached would be higher, as WaSH programs typically do J. Crocker et al. / Science of the Total Environment 601–602 (2017) 1075–1083 1079

Fig. 1. Program cost of four CLTS interventions in Ghana and Ethiopia, per household targeted. Abbreviations: NGO, non-governmental organization; CLTS, community-led total sanitation; NL, natural leader; HEW, health extension worker. All costs in this figure were borne by Plan. Costs are for the entire period of implementation. not reach all targeted beneficiaries.1 In Ghana, the cost of implementing here, which greatly impacted project costs, due to rough roads that ne- NGO-facilitated CLTS was $30.34 per household targeted, 70% of which cessitate expensive four-wheel drive vehicles, high fuel prices, and facil- was from facilitation. The addition of natural leader training raised costs itators spending significant time traveling to and from project villages. to $81.56 per household targeted. This substantial difference was most- However, these same villages may be most appropriate for CLTS, as ly due to expensive accommodation and meals at training venues, other WaSH projects often do not reach them, and CLTS is most effective which together were 70% of training costs. There were no low-cost ho- in these settings (Crocker et al., 2016a). tels capable of holding 80 natural leaders available. Training costs are subject to economies of scale, as venue rental and In Ethiopia, the cost of implementing HEW-facilitated CLTS was trainer pay would not increase with moderate increases in trainees. $19.21 per household targeted, which dropped to $14.15 for teacher- Total costs would also be lower where training venues are cheaper or facilitated CLTS. Management and training costs were lower for the closer to project villages. Management costs are also subject to econo- teacher-facilitated approach. Only one kebele in each region received mies of scale, provided that interventions are implemented with some HEW-facilitated CLTS. However, teachers were grouped together from degree of consistency across all project villages. Facilitation costs two kebeles in each region for training, which, on a per kebele basis, would be minimally if at all subject to economies of scale, as facilitation lowered management costs for planning training, venue rental cost, activities are proportional to the population targeted. and trainer costs. Facilitation cost was lower on average in teacher- facilitated CLTS kebeles because two of four teacher-facilitated kebeles 4.2. Local investments were not verified as ODF, eliminating the cost of ODF celebrations (not a desirable cost reduction). During NGO-facilitated CLTS in Ghana, local actors and community Program costs in Ghana were over four times those in Ethiopia on members invested time and money worth $7.93 per household average. Management and training costs in Ghana were approximately targeted, which rose to $22.36 where natural leaders were trained. double those in Ethiopia. Facilitation costs in Ghana were over ten times Household spending contributed approximately 75% in both cases, those in Ethiopia, because Plan trained local actors as facilitators in demonstrating that training natural leaders increased household invest- Ethiopia rather than leading facilitation themselves. In contrast, in ment in latrines. Local investments were much lower in Ethiopia: $3.41 Ghana, NGO staff led facilitation within villages. The program cost dif- per household targeted for HEW-facilitated CLTS, and $2.35 for teacher- ferences demonstrate how implementation arrangements can deter- facilitated CLTS. Most of the difference in local investments between mine costs, and thus how many people can be reached within a given countries was from very low spending on latrines in Ethiopia: $0.38 budget. Given that effectiveness also varies between interventions, we per household targeted, compared to $11.81 in Ghana, and some of would expect cost-effectiveness to vary even more than costs. the difference was due to lower value-of-time in Ethiopia where Despite dramatic differences in absolute costs between countries, wages are lower. Latrines in Ethiopia were built mostly of free, low- relative costs were similar in meaningful ways. Management was durability local materials. Local investments (by local actors and com- 26–28% of program cost for three of four interventions (excepting munity members) could also be call program outputs, as a beneficial CLTS with natural leader training in Ghana, in which accommodation and voluntary response to CLTS. and meals for training drove up total costs, reducing the management The three interventions in which Plan trained local actors demon- proportion). For the three interventions that included local actor train- strated that training leveraged local actors to facilitate or support ing, training cost was 56–61% of program cost. In both Ghana and CLTS. Where Plan trained local actors, for each hour that Plan spent on Ethiopia, over half of training cost came from accommodation and CLTS, local actors contributed 2.8 h in Ghana, and 4.7–6.8 h in meals, with trainee transport as the next largest portion. The relative Ethiopia. For the one intervention that did not include training local cost of management and training may reflect relative costs of other software-heavy behavior-change approaches. Table 4 The largest contributor to facilitation cost was transportation in both Breakdown of program costs for four CLTS interventions in Ghana and Ethiopia. fi Ghana and Ethiopia. CLTS projects often occur in remote areas and dif - Country Intervention Management Training Facilitation cult to access villages. This was the case for the interventions analyzed Ghana NGO CLTS 26% 4% 70% NGO CLTS + NL training 11% 58% 31% Ethiopia HEW CLTS 28% 56% 16% 1 Cost per population reached could be five times higher or more in this case, given the Teacher CLTS 27% 61% 11% low percent of the target population that changed from open defecation to latrine use (Crocker et al., 2017). However, converting to average cost per household reached can Abbreviations: NGO, non-governmental organization; CLTS, community-led total sanita- be misleading, as various outcomes can be chosen, and outcomes vary between settings. tion; NL, natural leader; HEW, health extension worker. 1080 J. Crocker et al. / Science of the Total Environment 601–602 (2017) 1075–1083

Table 5 Breakdown of local investments for four CLTS interventions in Ghana and Ethiopia.

Country Intervention Local Community Hired Purchased actors' members' time labor hardware time

Ghana NGO CLTS 5% 21% 26% 48% NGO CLTS + 7% 12% 20% 60% NL training Ethiopia HEW CLTS 35% 46% 0% 19% Teacher CLTS 43% 46% 0% 11%

Abbreviations: NGO, non-governmental organization; CLTS, community-led total sanitation; NL, natural leader; HEW, health extension worker.

Fig. 2. Components of program cost categories for community-led total sanitation fi interventions in Ghana and Ethiopia. All costs in this gure were borne by Plan. methods, and reported government-facilitated CLTS costing $1 per household reached in Ethiopia (costs per household targeted would be actors (NGO-facilitated CLTS in Ghana), local actors spent half as much even lower) (Sah and Negussie, 2009). The third TDC study reported time as Plan on CLTS. non-CLTS sanitation promotion as costing $2–45 per household Collectively, community members contributed the most time to targeted in three countries in South Asia, although methodological defi- CLTS, unsurprisingly given they are the targeted beneficiaries, and that ciencies are present, such as excluding paid government time despite CLTS is a participatory approach. Each hour Plan spent on CLTS led to government facilitating the project (Robinson, 2005). The fourth TDC community members contributing 5.9–7.5 h in Ghana, and 27–28 h in study reported software of non-CLTS sanitation promotion as costing Ethiopia. The higher ratios in Ethiopia do not represent a significantly $7–144 per household reached in six countries. Costs per household higher level of activity by local actors or communities; rather they rep- targeted were not reported, data collection was not described, and soft- resent Plan generating the same level of local activity with less of their ware cost components were unclear (Trémolet et al., 2010). own time. This efficiency is due to the interventions in Ethiopia focusing We found one BUC sanitation study, which reported that World on training local actors as facilitators. Bank-funded government-facilitated CLTS in Tanzania cost $30 per Individually, trained local actors spent 2.6%–12% FTE supporting household targeted, and $50 when hygiene promotion was included CLTS facilitation, with kebele leaders and HEWs in Ethiopia committing (Briceño and Chase, 2015). Because this was a BUC study, it may be the most time. While CLTS leverages investment of time by local actors, interpreted as more accurate and comprehensive than the TDC studies. it also burdens them. HEWs have many other job responsibilities. However, it used recall-based data, a non-representative sample of re- Kebele leaders are not compensated for time spent on CLTS. CLTS also spondents, and program costs were not disaggregated. leverages investment of time and money by community members. These prior studies exclude some cost categories, neglect local invest- The time-burden on community members was much lower than on ments, or use inappropriate TDC methods and non-representative re- local actors, and spending on latrines was voluntary. spondents, yielding inaccurate and underestimated costs. This makes comparison to our results difficult, as we used BUC and cover all program 4.3. Research in context costs and local investments. The low-end of the cost-range for all the TDC studies ($6, $1, $2, and $7) was dramatically lower than for the Four TDC studies have reported costs of sanitation promotion pro- one prior BUC study ($30), and for our study ($14). While this represents grams. The first TDC study reported that WaterAid CLTS programs in a small number of studies and covers different interventions implement- Bangladesh, Nepal, and Nigeria cost $6–84 per household targeted, but ed in different countries, it does support our hypothesis that TDC overhead costs were underreported and underestimated, the three methods will underestimate the cost of WASH programs, particularly countries had non-compatible financial tracking, and disaggregated those that are participatory or include behavior-change activities. The costs were “indicative” due to their TDC method (Evans et al., 2009). cost per household targeted range in our study overlaps with the range The second TDC study did not describe data collection or cost analysis of four of the five studies mentioned above.

Fig. 3. Local actor and community investments for four CLTS interventions in Ghana and Ethiopia, per household targeted. Abbreviations: NGO, non-governmental organization; CLTS, community-led total sanitation; NL, natural leader; HEW, health extension worker. All costs in this ﬁgure were borne by local actors and community members. Local actor and community member time was unpaid, and monetized using value-of-time assumptions. Hired labor and purchased hardware are household expenditures. J. Crocker et al. / Science of the Total Environment 601–602 (2017) 1075–1083 1081

Table 6 Time contributed to CLTS by different actors, and ratio of Plan's time to local actors' and community members' time, for four CLTS interventions in Ghana and Ethiopia.

Country Approach Full-time equivalent/10,000 peoplea Ratio of total Plan hours to:

Plan Local actors Community Local actor hoursb Community hoursc

Ghana NGO CLTS 1.4 0.7 8.4 1 to 0.5 1 to 5.9 NGO CLTS + NL training 2.0 5.5 15 1 to 2.8 1 to 7.5 Ethiopia HEW CLTS 0.7 3.3 19 1 to 4.7 1 to 27 Teacher CLTS 0.5 3.4 14 1 to 6.8 1 to 28

Abbreviations: NGO, non-governmental organization; CLTS, community-led total sanitation; NL, natural leader; HEW, health extension worker. a Full-time equivalent is 40 h per week. b Local actors includes local government and NLs in both countries, and kebele leaders, HEWs, and teachers in Ethiopia. c Community includes hired labor for latrine construction, and all other community activity related to CLTS.

When comparing our results to those of other studies, it is important estimated from survey data. Potential survey sampling error and bias to note that program costs and local investments are influenced by were minimized by a large, representative sample, and experienced social and political factors (in addition to being influenced by imple- contractors. Local investments were sensitive to value-of-time mentation arrangements, as discussed previously). In Ethiopia, every estimates, and may be underestimated given the 0.5 value-of-time to kebele is supposed to have a staffed health post and school, which minimum wage ratio we used. Transport cost calculations rely on as- enabled Plan approach to train facilitators, resulting in lower costs sumptions for vehicle depreciation, maintenance, travel time, and driv- than in Ghana. Ethiopia has a weaker economy, lower GDP, and lower ing speed. Assumptions were based on real data for vehicles used, and wages than in Ghana, which resulted in lower costs for Plan, and AAA travel cost models, which may underestimate maintenance and value-of-time for local actors and community members, which made fuel costs for travel on rough roads. A sensitivity analysis was conducted both program costs and local investments lower than in Ghana. This on value-of-time and transportation parameters to test their robust- may also explain the lower expenditures on latrines in Ethiopia. ness. For contracted work (LNGO facilitation, and district government monitoring), cost allocation to management, salary, and transport cate- 4.4. Contributions gories was based on submitted budgets, which may deviate from exact expenditures by category. However, the total contract costs are accu- This is the first study to provide comprehensive, accurately tracked, rate, as they reflect payments. Local investments likely continued after disaggregated costs for any WaSH behavior-change program. This study cost tracking as part of this study stopped (in contrast to program provides insight into the cost of project management, training, and facil- costs, which did not continue). itation, all of which are ubiquitous among participatory behavior- change public health projects, making these findings relevant beyond 5. Conclusions WaSH. The evidence provided by this study is particularly important given that the SDGs emphasize capacity building, local participation, This is the first study to present comprehensive, accurate, disaggre- and public finance to leverage other investments such as local actor gated costs of a WaSH behavior-change program. These findings can and community investments (UN General Assembly, 2015), all of add value to WaSH policy and planning discussions, and can be incorpo- which are present in the interventions analyzed here. rated into cost-effectiveness research as the first reliable cost figures for We developed data collection tools and a framework for cost analy- a CLTS intervention. Evidence on process and costs is an important re- sis of complex programs that are implemented by multiple organiza- source for policy and funding decisions, program design and manage- tions, are flexible and adaptable, involve cross subsidies, or include ment, cost-effectiveness and cost-benefit studies, modeling program local investments. As these program characteristics are common scale up, and evaluating and comparing different interventions. beyond WaSH, these tools are relevant for other public health programs. This study offers a robust method and process for tracking and calcu- For example, interventions to influence use of clean cookstoves, postna- lating costs, which can be followed and further developed to build evi- tal health behaviors, and HIV prevention behaviors are all common dence on a broad range of WaSH interventions that is consistent and public health behavior-change interventions that contain the above accurate. The multi-site, multi-intervention research approach used in characteristics. this study is important for understanding how fi ndings would transfer to other programs and settings. 4.5. Limitations

Our findings are context specific, as costs vary between settings and Competing interests interventions. However, this study covered four interventions across five regions in two countries, and presents disaggregated results, to DS is employed by Plan International USA, and oversaw the inter- show how costs vary by setting and intervention, and how implementa- ventions evaluated. The other authors declare no real or potential con- tion activities drive cost variation. Some local investments were flicts of interest.

Table 7 Time spent on CLTS implementation by local actors and community members.

Country Approach Average hours per-person per-weeka (FTEb)

Govt. KLs HEWs Teachers NLs Community

Ghana NGO CLTS 2.5a (6.2%)b –––0.18 (0.45%) 0.04 (0.09%) NGO CLTS + NL training –––1.5 (3.7%) 0.07 (0.16%) Ethiopia HEW CLTS 1.1 (2.6%) 2.9 (7.2%) 3.4 (8.5%) –– 0.09 (0.23%) Teacher CLTS 4.6 (12%) – 2.2 (5.4%) – 0.07 (0.17%)

Abbreviations: NGO, non-governmental organization; CLTS, community-led total sanitation; KL, kebele leader; NL, natural leader; HEW, health extension worker; FTE, full-time equivalent. a Average hours-per-week spent on CLTS per person. b Full-time equivalent is 40 h per week. 1082 J. Crocker et al. / Science of the Total Environment 601–602 (2017) 1075–1083

Fig. 4. Hours spent on CLTS by different actors per month, Ghana and Ethiopia. Abbreviations: CLTS, community-led total sanitation. Plan activities ended in month 20 in Ghana, and in month 13 in Ethiopia. Local actor and community member activity likely continued after Plan activities ended, but was not tracked.

Acknowledgements Crocker, J., Saywell, D., Bartram, J., 2017. Sustainability of community-led total sanitation outcomes: evidence from Ethiopia and Ghana. Int. J. Hyg. Environ. Health http://dx. doi.org/10.1016/j.ijheh.2017.02.011. Funding: This research is a result of a subagreement to the University Dangour, A.D., Watson, L., Cumming, O., Boisson, S., Che, Y., Velleman, Y., Cavill, S., Allen, of North Carolina at Chapel Hill from Plan International USA, which re- E., Uauy, R., 2013. Interventions to improve water quality and supply, sanitation and hygiene practices, and their effects on the nutritional status of children (review). ceived a grant from the Bill & Melinda Gates Foundation (OPP1028953). Cochrane Libr (99 pp.). http://dx.doi.org/10.7326/M13-2199. A National Institute of Environmental Health Sciences Training Grant Dickinson, K.L., Patil, S.R., Pattanayak, S.K., Poulos, C., Yang, J.-C., 2015. Nature's call: im- (T32ES007018) supported the primary author. Neither funder influenced pacts of sanitation choices in Orissa, India. Econ. Dev. Cult. Chang. 64:1–29. http:// study design, execution, or article preparation. dx.doi.org/10.1086/682958. Evans, B., Jones, H., Colin, J., Robinson, A., 2009. Sustainability and Equity Aspects of Total The authors would like to thank: the study participants; Plan Inter- Sanitation Programmes: A Study of Recent WaterAid-supported Programmes in national USA for grant management; Plan International Ghana and Three Countries: Global Synthesis Report. WaterAid, London. Ethiopia for implementation; and Subhrendu Pattanayak for reviewing Garber, A.M., Phelps, C.E., 1997. Economic foundations of cost-effectiveness analysis. J. Health Econ. 16, 1–31. the draft manuscript. Garn, J.V., Sclar, G.D., Freeman, M.C., Penakalapati, G., Alexander, K.T., Brooks, P., Rehfuess, E.A., Boisson, S., Medlicott, K.O., Clasen, T.F., 2017. The impact of sanitation interven- Appendix A. Supplementary data tions on latrine coverage and latrine use: a systematic review and meta analysis. Int. J. Hyg. Environ. Health 220:329–340. http://dx.doi.org/10.1016/j.ijheh.2016.10.001. Haller, L., Hutton, G., Bartram, J., 2007. Estimating the costs and health benefits of water Supplementary data to this article can be found online at http://dx. and sanitation improvements at global level. J. Water Health 5:467–480. http://dx. doi.org/10.1016/j.scitotenv.2017.05.279. doi.org/10.2166/wh.2007.008. Hunter, P.R., Pond, K., Jagals, P., Cameron, J., 2009. An assessment of the costs and benefits of interventions aimed at improving rural community water supplies in developed References countries. Sci. Total Environ. 407:3681–3685. http://dx.doi.org/10.1016/j.scitotenv. 2009.03.013. AAA, 2015. Your Driving Costs: How Much Are You Really Paying to Drive? Hutton, G., Varughese, M., 2016. The Costs of Meeting the 2030 Sustainable Development Adam, T., Evans, D.B., Koopmanschap, M.A., 2003. Cost-effectiveness analysis: can we re- Goal Targets on Drinking Water, Sanitation, and Hygiene. fi duce variability in costing methods? Int. J. Technol. Assess. Health Care 19, 407–420. Hutton, G., Haller, L., Bartram, J., 2007. Global cost-bene t analysis of water supply and Borghi, J., Guinness, L., Ouedraogo, J., Curtis, V.A., 2002. Is hygiene promotion cost- sanitation interventions. J. Water Health 5:481–502. http://dx.doi.org/10.2166/wh. effective? A case study in Burkina Faso. Trop. Med. Int. Health 7, 960–969. 2007.009. Briceño, B., Chase, C., 2015. Cost-efficiency of rural sanitation promotion: activity-based Institute of Development Studies, 2016. The CLTS approach. [WWW Document]. URL. costing and experimental evidence from Tanzania. J. Dev. Eff. 1–12. http://dx.doi. http://www.communityledtotalsanitation.org/page/clts-approach (accessed 9.30. org/10.1080/19439342.2015.1105848. 16). Burr, P., Fonseca, C., 2011. Applying the Life-cycle Costs Approach to Sanitation: Costs and Kar, K., Chambers, R., 2008. Handbook on Community-led Total Sanitation. Plan UK and Service Levels in Andhra Pradesh (India). Burkina Faso, Ghana and Mozambique. Institute of Development Studies. Carey, K., Burgess, J.F., 2000. Hospital costing: experience from the VHA. Financ. Account. Plan International Ethiopia, 2015. Teacher-facilitated community-led total sanitation: im- Manag. 16, 289–308. plementation narrative. [WWW Document]. URL. https://waterinstitute.unc.edu/ fi Chapko, M.K., Liu, C.-F., Perkins, M., Li, Y.-F., Fortney, J.C., Maciejewski, M.L., 2009. les/2015/12/clts-gates-ethiopia-2015.pdf. Equivelence of two healthcare costing methods: bottom-up and top-down. Health Plan International Ghana, 2015. CLTS with capacity building for natural leaders: imple- fi Econ. 18:1188–1201. http://dx.doi.org/10.1002/hec.1422. mentation narrative. [WWW Document]. URL. https://waterinstitute.unc.edu/ les/ Crocker, J., Abodoo, E., Asamani, D., Domapielle, W., Gyapong, B., Bartram, J., 2016a. Im- 2015/12/clts-gates-ghana-2015.pdf. pact evaluation of training natural leaders during a community-led total sanitation Prüss-Ustün, A., Bartram, J., Clasen, T.F., Colford, J.M., Cumming, O., Curtis, V.A., Bonjour, S., intervention: a cluster-randomized field trial in Ghana. Environ. Sci. Technol. 50: Dangour, A.D., De France, J., Fewtrell, L., Freeman, M.C., Gordon, B., Hunter, P.R., John- 8867–8875. http://dx.doi.org/10.1021/acs.est.6b01557. ston, R.B., Mathers, C., Mäusezahl, D., Medlicott, K., Neira, M., Stocks, M., Wolf, J., Crocker, J., Geremew, A., Atalie, F., Yetie, M., Bartram, J., 2016b. Teachers and sanitation Cairncross, S., 2014. Burden of disease from inadequate water, sanitation and hygiene promotion: an assessment of community-led total sanitation in Ethiopia. Environ. in low- and middle-income settings: a retrospective analysis of data from 145 coun- – Sci. Technol. 50:6517–6525. http://dx.doi.org/10.1021/acs.est.6b01021. tries. Trop. Med. Int. Health 19:894 905. http://dx.doi.org/10.1111/tmi.12329. J. Crocker et al. / Science of the Total Environment 601–602 (2017) 1075–1083 1083

Robinson, A., 2005. Scaling-up Rural Sanitation in South Asia: Lessons Learned From UN General Assembly, 2015. Draft Outcome Document of the United Nations Summit for Bangladesh, India, and Pakistan. Water and Sanitation Program-South Asia, New Delhi. the Adoption of the Post-2015 Development Agenda. http://dx.doi.org/10.1093/ Sah, S., Negussie, A., 2009. Community led total sanitation (CLTS): addressing the chal- oxfordhb/9780199560103.003.0005. lenges of scale and sustainability in rural Africa. Desalination 248, 666–672. Venkataramanan, V., 2012. Testing CLTS Approaches for Scalability: Systematic Literature Spears, D., 2013. How much international variation in child height can sanitation explain? Review. UNC and Plan International USA, Chapel Hill, NC. (No. 6351). World Bank Policy Research Working Paper. WPS 6351 Washington, DC. Venkataramanan, V., 2016. CLTS Learning Series: Lessons From CLTS Implementation in The World Bank, 2017. GDP per capita (current US$). [WWW Document]. World Bank Seven Countries. UNC and Plan International USA, Chapel Hill, NC. Natl. Accounts Data, OECD Natl. Accounts Data Files URL. http://data.worldbank. Whittington, D., Jeuland, M.A., Barker, K., Yuen, Y., 2012. Setting priorities, targeting sub- org/indicator/NY.GDP.PCAP.CD (accessed 4.18.17). sides among water, sanitation, and preventative health interventions in developing Trémolet, S., Kolsky, P., Perez, E., 2010. Financing On-site Sanitation for the Poor: A Six countries. World Dev. 40:1546–1568. http://dx.doi.org/10.1016/j.worlddev.2012.03. Country Comparative Review and Analysis. World Bank, Water and Sanitation Pro- 004. gram, Washington, DC.